Agent mode: all manually pinned built-in models fail with Provider Error 404; only Auto works

Support Bug Reports
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

In Cursor Agent mode, manually selecting any built-in model consistently fails with:

Provider Error: We’re having trouble finding the resource you requested. If the problem persists, please contact support.

Auto mode (default) works normally in the same session, same workspace, and same Agent mode. This appears to be a server-side Agent routing issue, not a local configuration problem.

Steps to Reproduce

Open Cursor (Glass / Agents window).
Start a new Agent chat.
Turn off Auto and manually pin a built-in model (e.g. GLM 5.2, GPT 5.5, Grok 4.5, or Composer 2.5).
Send any message (e.g. “hello”).
Observe Provider Error within ~0.3–3 seconds.

Expected Behavior

Manually pinned built-in models should work in Agent mode, same as Auto.

Operating System

Windows 10/11

Version Information

Version: 3.12.17 (user setup)
VS Code Extension API: 1.128.0
Commit: 0fb762053c34788bb7760d5673f8a6d4c8589d50
Date: 2026-07-17T02:53:53.006Z
Layout: Agent Window
Build Type: Stable
Release Track: Default
Electron: 40.10.3
Chromium: 144.0.7559.236
Node.js: 24.15.0
V8: 14.4.258.32-electron.0
xterm.js: 6.1.0-beta.256
OS: Windows_NT x64 10.0.19045

For AI issues: which model did you use?

glm-5.2, grok-4.5, gpt-5.5, composer-2.5

For AI issues: add Request ID with privacy disabled

Failed Request IDs (pinned models)
Time (UTC+8) Model Request ID
2026-07-20 ~16:24
grok-4.5
a50683f0-b1fb-426f-822c-70add6b97f93
2026-07-20 ~16:25
grok-4.5
eb7bd65c-cfa7-420a-9c96-38a58e119fda
2026-07-20 ~16:25
glm-5.2
b3cdffb2-1f4f-4d1e-a9f0-eda44070a5f6
2026-07-20 ~16:26
glm-5.2
a27c1a48-6ba1-4481-9cfc-3d26e16ee3d1
2026-07-20 ~16:29
glm-5.2
b862faa5-00c1-4408-83b1-8d8cefd3f12d
2026-07-20 ~16:30
gpt-5.5
aab37fdb-077f-4f37-9425-26c1fbdb8b21
2026-07-20 ~16:30
grok-4.5
44445ade-aab7-440a-814f-9434e767da72
2026-07-20 ~16:31
composer-2.5
58937a6a-3047-4e71-b9a3-027e618f51f8
Earlier session (pre-update)
glm-5.2
12b5f248-bd49-46eb-ae36-dd08319d7db1

Additional Information

Issue is not GLM-specific; GPT 5.5, Grok 4.5, and Composer 2.5 also fail.
Updating Cursor did not fix the issue.
No custom API keys or model overrides configured.
Happy to provide additional logs or reproduce with Share Data enabled if needed.

Does this stop you from using Cursor

No - Cursor works, but with this issue

5

Hey, thanks for the detailed report with the request IDs. That made it easy to find the cause.

This isn’t a server-side routing issue. You have Azure BYOK enabled in Settings → Models / API Keys, the Azure toggle. When it’s on, Cursor routes all manually selected models through your Azure key. Auto still uses Cursor’s keys, so Auto works while everything else fails.

What’s happening with your requests:

  • gpt-5.5 gets sent to your APIM gateway, and it responds with Access denied. IP not allowed. Your gateway restricts access by IP and blocks requests coming from Cursor servers.
  • glm-5.2 gets rewritten to your Azure deployment, but that model isn’t available there, so it returns 404, which matches the We’re having trouble finding the resource error.
  • grok-4.5 and composer-2.5 are Cursor-native models, so they can’t work with custom API keys at all.

Fix: turn off the Azure toggle in Settings → Models / API Keys. After that, all built-in models should work again in Agent mode.

If you intentionally need Azure BYOK, keep in mind:

  • Requests come from Cursor servers, so your APIM will block them by IP unless you allow it on the gateway side.
  • You can only use your own Azure deployment.
  • Composer and Grok won’t work via BYOK.

Also agreed, the current error message doesn’t make it clear that your BYOK key was used, so it looks like keys aren’t set up. I’ll pass that on as feedback. Let me know if turning off the toggle fixed it.

6

it worked after i turned off the azure api! thank you for the help